统计与数据科学学院知识库

农业是国民经济的基础，而大豆在我国农业中占据着重要的位置，大豆市场的波动影响着我国农业的发展乃至社会经济的稳定；期货在经济发展过程中能够有效地规避风险，提高市场的透明度。因此，对大豆期货价格的研究具有重要意义。

本文将数据分解方法与机器学习方法相结合，基于“分解-集成”框架下的大豆期货价格预测分为以下两个部分：

第一部分是基于分解去噪框架的大豆期货价格预测研究，对大豆期货价格进行分解去噪，并引入与大豆紧密相关的豆粕和玉米的期货价格信息对大豆期货价格进行预测。期货价格序列中含有噪声，这可能会掩盖豆粕期货价格与玉米期货价格对大豆期货价格的影响，因此在本文中对引入的豆粕期货价格与玉米期货价格进行去噪处理。结果表明，分解去噪可以有效地提升预测效果，而引入豆粕期货和玉米期货的信息能进一步提升模型的预测性能。

第二部分是基于“分解-重构-集成”框架的大豆期货价格预测研究，主要研究不同的重构方法对大豆期货价格预测的影响，并在此基础上引入其它期货市场的信息。本文的重构方法包括考虑复杂性、周期性和与原始序列的相关性三种数据特征的单特征重构与多特征重构。本文引入豆粕期货与玉米期货的价格信息，对豆粕期货价格与玉米期货价格进行分解重构，并对豆粕期货价格和玉米期货价格的重构序列与大豆期货重构序列进行相关性分析，从而筛选出预测大豆期货重构序列的输入变量。实证结果表明多特征重构方法优于单特征重构方法，引入其它期货市场信息能够有效提升模型的预测性能。

本文的研究层层深入，所提出的预测模型在大豆期货价格预测方面的良好表现为大豆期货价格预测提供了一种新的思路。

Agriculture is the foundation of the national economy, and soybeans occupy an important position in China's agriculture. The fluctuations in the soybean market affect the development of China's agriculture and even the stability of the social economy; Futures can effectively avoid risks and improve market transparency in the process of economic development. Therefore, the study of soybean futures prices is of great significance.

This thesis combines data decomposition methods with machine learning methods, and the soybean futures price prediction based on the "decomposition integration" framework is divided into the following two parts:

The first part is a study on soybean futures price prediction based on a decomposition denoising framework, which decomposes and denoises soybean futures prices and introduces futures price information of soybean meal and corn closely related to soybeans to predict soybean futures prices. The futures price sequence contains noise, which may mask the impact of soybean meal futures prices and corn futures prices on soybean futures prices. Therefore, in this thesis, the introduced soybean meal futures prices and corn futures prices are denoised. The results indicate that decomposition denoising can effectively improve the prediction performance, while introducing information from soybean meal futures and corn futures can further improve the predictive performance of the model.

The second part is a study on soybean futures price prediction based on the "decomposition reconstruction integration" framework, mainly studying the impact of different reconstruction methods on soybean futures price prediction, and introducing information from other futures markets on this basis. The reconstruction method in this thesis includes single feature reconstruction and multi feature reconstruction considering three data features: complexity, periodicity, and correlation with the original sequence. This thesis introduces the price information of soybean meal futures and corn futures, decomposes and reconstructs the prices of soybean meal futures and corn futures, and analyzes the correlation between the reconstructed sequence of soybean meal futures prices and corn futures prices and the reconstructed sequence of soybean futures, thereby screening the input variables for predicting the reconstructed sequence of soybean futures. Empirical evidence shows that the multi feature reconstruction method is superior to the single feature reconstruction method, and introducing other futures market information can effectively improve the predictive performance of the model.

The research in this thesis is in-depth layer by layer, and the good performance of the proposed prediction model in predicting soybean futures prices provides a new approach for predicting soybean futures prices.